Rtx 3080 lower quality capacitor Issue

[Adzassin]

Have been doing a bit of testing with my Gigabyte gaming and found that it doesnt like anything past +100 core and gets unstable upwards of 2055 boost. Will continue more real world stuff with the overclock to see if it produces any instability but to be honest im not that unhappy if it can be stable at 2000.

 

[SeeNoWeevil]

they allow their cards to boost and without adacute testing, they fail due to not being able to handle the speeds they're allowing their cards to go to.

You're agreeing with me.

 

[cutterslade]

Here is another post of TiN
https://www.techpowerup.com/forums/...d-capacitor-choice.272591/page-8#post-4357145
mentioning why the capacitors below the GPU might not be the problem.

Here is a post of a guy testing a non-premium Gigabyte 3080 card over 2000 MHz without problems although that card has the mentioned POSCAP capacitors
https://new.reddit.com/r/nvidia/com..._crashes_and_instabilities/g6pgp3t/?context=3

I think we need to wait a bit and drink tea.

 

[cutterslade]

And here, a guy is mentioning that also Asus cards have problems
https://twitter.com/HardwareUnboxed/status/1309659834468298753

When we look at the silence of Nvidia and the board partners, the root cause may not have been found, yet.

 

[Toekiller]

Somehow many FPS are people losing if they can run at 1900 and not 2100?

 

[chrcoluk]

So, for the non overclockers amongst us, i assume no problems

As long as you dont buy an OC card, which I wouldnt anyway as its a small boost for too much price premium..

 

[Nasher]

Somehow many FPS are people losing if they can run at 1900 and not 2100?

Not even a noticable amount.

 

[cutterslade]

Its not about OC I would say. The question is why those problem cards are suddenly peaking over 2000 MHz on boost. Why does the GPU think its ok to go that far?

 

[Grim5]

Somehow many FPS are people losing if they can run at 1900 and not 2100?

3-5% for every 100mhz

 

[ubersonic]

Yep, blaming Nvidia is a cop out.

O.o

Nvidia gave AIBs a reference design/spec with a design flaw.

Nvidia rushed the launch, thus denying AIBs the time needed to spot the design flaw.

Who else is there to blame?

 

[Dave2150]

Not even a noticable amount.

As Grim said, a 5-10% difference. IMO This is noticeable, especially when at 4k/VR where you really need the extra frames.

 

[SkinSpotlights]

As Grim said, a 5-10% difference. IMO This is noticeable, especially when at 4k/VR where you really need the extra frames.

Honestly when I get it I'm probably going to do the underclock/undervolt so I drop 3fps for 100 watts less power consumption. If I was running 4k I 100% wouldn't, like you said it makes more of a difference when at 4k and VR if the headset is a large resolution.

 

[cutterslade]

O.o

Nvidia gave AIBs a reference design/spec with a design flaw.

Nvidia rushed the launch, thus denying AIBs the time needed to spot the design flaw.

Who else is there to blame?

Extreme capitalism, the constant chase for max winnings for no matter what.

 

[chrcoluk]

Right now, not really.

• We know the review and early shipped units of the Trio has a 5-1 arrangement of the larger capcitors to 10 smaller capacitors.
• We know EVGA said they found that for their FTW3 card they needed a 4-2 arrangement, but that their XC3 was fine with 5-1.

• We don't know if the current Trios are shipping the same as the initial ones, or if MSI picked up on the issue and changed at some point (which might explain the nearly 0 numbers of them that seem to be out in the wild).
  
• We don't know what exact difference between the XC3 and FTW3 means one was fine and the other needed more, and how that translates to the Trio.
  
• We don't know that the capacitors are the only issue, since some people are reporting to have similar problems with suposidly "good" layout ASUS Cards and FE Cards

Its a developing issue, and while it seems fairly certain that the capacitors are part of what is causing it, it might be more than that.

Personally I suspect it'll turn out to either be the 30% "ok" Bin 0 chips or the 10% "really good" Bin 2 chips that are susceptible this issue (either because they can't stay stable at higher boosts, or can boost higher and draw more current which the power delivery can't handle). EVGA just said they found the problem in testing, not that it affected all the cards, and from what we've heard AIB's didn't have time to actually Bin the chips themselves before starting manufacturing.

Interesting to learn that nvidia have ditched their pre-binning.

He also reveals what we already know but never confirmed that press samples are typically bin 2.

 

[killari]

No one has promised that the GPU should run at ~2000+mhz clock. Each vendor has a stated base and boost clock, which is a specification that it looks like they do meet, as well as the opportunistic boosting algorithm which is advertised to try (but no promises) to push things further. It sounds like the algorithm is just a little too gung-ho. If a software update was released that bumped clocks down a little, so long as the card is stable at boost clocks under the vendor's definition of "normal operating conditions" the marketing promise is still kept.

The outrage just sounds like unjustified/unreasonable expectations. No one has stated that people should expect any clock speed beyond boost (GPU boost is a promise that the card will try, and try it does), let alone ~2000+mhz (which in itself is a rather arbitrary round number) so why are they?

Also from what I've read so far, we don't know that a problem lies with the caps, its purely speculation. EVGA made some commentary (which is in no way comprehensive) and the more credible Youtube "experts" have argued that no cap type is potentially worse than the other, they both have pros and cons.

 

[chrcoluk]

Did he disclose these crashes in his review, if not why not?

Its kind of like the 3.5 gig fiasco again where the reviewers only spoke out after everything went public

 

[Grim5]

Did he disclose these crashes in his review, if not why not?

Its kind of like the 3.5 gig fiasco again where the reviewers only spoke out after everything went public

He did not

most reviewers failed to mention their cards were crashing, it's quite deceptive if you ask me

 

[MagicBoy]

People who have spent that much money on new hardware may have to make their cards under perform to prevent a crash to desktop. Is this worse than the scandal where cards were advertised as 4GB but there was only 3.5GB on the board?

That's false. The 970 had 4GB, however the last 512MB was slower to access due to the cut down number of shaders and the memory controller design compared to the 980.

 

[SkinSpotlights]

He did not

most reviewers failed to mention their cards were crashing, it's quite deceptive if you ask me

Very disappointing for them to do that, curious what the excuse will be, "Oh we were told by the manufacturer its a pre-production model issue".

Which other reviewers failed to disclose their cards crashing?

 

[cutterslade]